Characterizing the capability of public DEMs for mapping global floodplain bathymetry

Digital elevation models (DEMs) play a significant role in floodplain management, including water inundation boundary delineation, flood modeling and forecasting, and risk assessment. The impact of spatio-temporal resolution and vertical accuracy of DEMs on flood modeling has been widely concerned in previous studies. However, due to variations in data acquisition season and geographical coverage of each DEM, it still remains unclear how much variation exists in the extent of exposed topography provided by different DEMs over the global floodplains. Furthermore, how to select and jointly use the various DEM data sources for achieving the maximum level of exposed topography coverage also requires to be further explored. This study selected three DEMs (SRTM DEM, ASTER DEM, and FABDEM) to evaluate their capability of offering the topographic exposure extent within global floodplains by proposing the topographic exposure index (TEI). The results show that the global average TEI is highest for SRTM DEM, followed by ASTER DEM, while FABDEM has the lowest average TEI. The difference in TEI distribution between FABDEM and the other two DEMs is the most significant. Moreover, the TEI of these three DEMs has remarkably inconsistent performances in spatial sub-zones. In regions such as southern China, central India, and Brazil, the TEI of SRTM DEM and ASTER DEM is significantly higher compared to FABDEM, largely influenced by human activities such as dam-induced reservoir water impoundment areas and aquacultures. In contrast, FABDEM shows significantly higher TEI than the other two DEMs in Canada and central Siberia, as well as in regions with saline lakes (e.g., the Great Salt Lake, Lake Razzaza), predominantly influenced by climatic and hydrologic features. The joint use of these three DEM datasets with appropriate fusing strategy may expand the exposed topographic area of global floodplains, at an equivalent magnitude of the Nile floodplain area (2.67 x 105 km2).